A high-resolution subsurface model provides vital information both for scientific purposes of understanding the earth, and for engineering purposes including hydrocarbon exploration and monitoring, and CO2 sequestration monitoring. Seismic methods extract spatial distributions of physical properties of the earth from seismic data generated by controlled sources (active seismic) or by earthquakes (passive seismic). Seismic full waveform inversion ambitiously aims to exploit entire recorded wavefields, and achieves unprecedentedly high resolution when compared with conventional methods which focus on arrival times of distinct waves. Theoretical and algorithmic improvements are key to further enhance the applicability of full waveform inversion.
Full waveform inversion ideally explores an optimum model of subsurface elastic properties by fitting “full” waveforms consisting of transmitted, reflected and multiple-scattered arrivals. Due to the increase in the amount of information used, full waveform inversion achieves a remarkable resolution, and provides additional complementary information to conventional processing results. However the current practices in exploration seismology typically limit to fit early arrivals, and to search for a P-wave velocity. Inversion methods to truly exploit full wavefields and to expand to multi-parameter are currently under investigation together with Winthrop Professor David Lumley.
Full waveform inversion is a challenging inverse problem both from inverse-theory and computational points of view. The optimum model is defined as a minimum of an objective function constructed from differences between observed wavefield and modelled wavefields. Wavefield modelling is conducted by numerically solving the wave equation, and this computational expensive operation requires the use of high performance computers. In addition, the inverse problem is highly non-linear and non-unique. Therefore the inversion is sensitive to a starting model, and the convergence to the minimum tends to be slow. This is also a part of the reason of the limitations in the current practices described above. Dr Kamei is developing robust inversion theory and strategies to improve these issues, while maintaining the computational cost at a reasonable level.
Dr Kamei is also striving to expand the applications of full waveform inversion to passive seismic monitoring. Full waveform inversion potentially yields a subsurface model to improve the estimates of earthquake locations, and also to evaluate changes in reservoirs due to induced micro-seismicity from hydrocarbon production and CO2 sequestration. Both are important information to characterize changes in reservoirs related to fluid migration.